Crunchbase alternative: U.S. company lists as a free file

A free account gives company profiles, search results, AI-powered research and Crunchbase MCP, with usage limits, says the same page. Crunchbase's Pro page counts 4M+ private companies. At 2,000 rows a month, the first list below, 22,246 companies, is 12 months of Pro's exports, and 23 searches of 1,000 results.

The year founded is the one on the company's LinkedIn page, as the company gave it. The free file holds companies of 11 to 500 employees; a startup of fewer is in the 50M+ file (below). Every column, with its fill rate: the free US B2B leads dataset. A free sample, no sign-up: 10,000 U.S. companies drawn from the free list, companies only, as CSV and Parquet, CC BY 4.0. See the sample.

Unzip the free file into us_10m/ and pip install duckdb. Each query below ran as printed on the October 2026 file, on a laptop, in under a second. First, the software, internet and IT companies founded in 2020 or later:

import duckdb

# U.S. software, internet and IT companies founded in 2020 or later
duckdb.sql("""
  copy (
    select NAME, URL, LINKEDIN_INDUSTRY, EMPLOYEE_COUNT_RANGE, FOUNDED, HQ_CITY, HQ_STATE_CODE, LINKEDIN_URL
    from 'us_10m/us_smb_mid_market_companies.parquet'
    where FOUNDED >= 2020
      and LINKEDIN_INDUSTRY in ('Software Development', 'Technology; Information and Internet', 'IT Services and IT Consulting')
  ) to 'startups.csv'
""")

It writes 22,246 companies. Add and HQ_STATE_CODE = 'CA' for California's 4,057; change the year or the industries for another list.

The companies that say they work on AI, in any industry:

# Companies founded in 2020 or later whose LinkedIn tagline or description mentions AI
duckdb.sql(r"""
  copy (
    select NAME, URL, LINKEDIN_INDUSTRY, EMPLOYEE_COUNT_RANGE, FOUNDED, HQ_CITY, HQ_STATE_CODE, HEADLINE, LINKEDIN_URL
    from 'us_10m/us_smb_mid_market_companies.parquet'
    where FOUNDED >= 2020
      and (regexp_matches(concat_ws(' ', HEADLINE, ABOUT), '\bAI\b|\bLLMs?\b')
        or regexp_matches(concat_ws(' ', HEADLINE, ABOUT), 'artificial intelligence|machine learning', 'i'))
  ) to 'ai_companies.csv'
""")

It writes 17,354 companies. It reads what each company wrote on LinkedIn, so a company that only mentions AI is in it too: read the HEADLINE column before you write to them.

Then the people who run the first list's companies, with an email or a mobile:

# The founders and CEOs of the first list's companies, with an email or a mobile
duckdb.sql(r"""
  copy (
    select p.FULL_NAME, p.CURRENT_JOB_TITLE, c.NAME as COMPANY, c.URL as WEBSITE, c.FOUNDED, p.EMAIL, p.MOBILE_PHONE, p.LINKEDIN_URL
    from 'us_10m/us_smb_mid_market_persons.parquet' p
    join 'us_10m/us_smb_mid_market_companies.parquet' c on c.LINKEDIN_ID = p.CURRENT_JOB_COMPANY_LINKEDIN_ID
    where c.FOUNDED >= 2020
      and c.LINKEDIN_INDUSTRY in ('Software Development', 'Technology; Information and Internet', 'IT Services and IT Consulting')
      and regexp_matches(p.CURRENT_JOB_TITLE, '\b(founder|co-founder|cofounder|ceo|chief executive officer)\b', 'i')
      -- an executive assistant to the CEO isn't a CEO
      and not regexp_matches(p.CURRENT_JOB_TITLE, '\bassistant\b.*\b(to|for)\b|\bexecutive assistant\b', 'i')
      and (p.EMAIL is not null or p.MOBILE_PHONE is not null)
  ) to 'startup_founders.csv'
""")

It writes 1,740 founders and CEOs at 1,550 of the 22,246 companies. Any other title: email lists by job title. A list by state, industry and size, every founding year: free list of US companies.

50M+ U.S. B2B: 43.6M people and 6.7M companies in one 9.6 GB zip, a new one every month. Its companies of every size include 732,854 founded in 2020 or later, 592,782 of them of 1 to 10 employees, and 93,886 in software, internet and IT. The queries above run on it as they are, with its files' names. Add $50 to your account: you get $50 of API PLUS the flat file. Or invite 3 people who sign up.

• Funding rounds, investors, valuations, acquisitions and IPOs.
• Growth and heat scores, and forecasts of funding, acquisitions and IPOs.
• Companies beyond the U.S.: the files hold U.S. companies only.
• Its AI search, Scout, its Chrome extension and its CRM integrations on Business.

For a growth signal from the file itself, the companies that hired in 3 of the last 4 quarters: why SQL is superior when building your list.

Datacircle sells data, not research software. Its API answers the LinkedIn profile behind a URL, the person's whole profile, at $0.00125 a profile found: LinkedIn profile API. The company data is in the file, on your dashboard once you sign up with your work email, with the $5 credit and your API key.